CTO Tech Talk: Comparing Elasticsearch and Rockset Streaming Ingest and Query Performance

In this tech talk, Rockset CTO Dhruba Borthakur and founding engineer and architect Igor Canadi go under the hood of Elasticsearch and Rockset to compare and contrast their performance and architectures for workloads spanning from real-time analytics to text search and vector search.

They’ll describe key architectural design differences including:

- Apache Lucene vs. RocksDB: Elasticsearch uses Lucene, an open-source immutable search engine library written in Java, whereas Rockset uses RocksDB, an open-source mutable key-value store developed at Meta.

- Tightly Coupled vs. Decoupled Architectures: Elasticsearch tightly couples compute and storage to achieve fast performance. In contrast, Rockset adopts a loosely coupled architecture with compute-storage and compute-compute separation for isolation and resource efficiency.

- Search vs. Converged Indexing: Both Elasticsearch and Rockset rely on indexing for high-performance queries. Elasticsearch uses a search index which is ideal for needle-in-the-haystack queries commonly found in text search and log analytics. Rockset has a Converged Index which combines a search index, columnar store and row store for search, aggregations and joins in real-time analytics.

The talk will conclude by sharing streaming ingestion and query performance results from real-life customer applications.